Using AI study tools becomes safer when students understand the everyday words public, private, personal, confidential, permission, consent and share. These terms decide what belongs in a prompt—not merely how politely the prompt is written.
The practical answer: before pasting a worksheet, group-chat message, photograph or class document into an AI tool, ask who supplied it, whether it identifies anybody, how sensitive it is and whether that material is authorised for that particular tool. When uncertain, use an invented example or a teacher-approved extract instead. Students can still learn English, Mathematics and Science without exposing classmates’ information.
Did you know? A document can reveal someone’s identity even without their full name. A class photograph, a distinctive quotation, a group-chat screenshot or several small clues together might identify a real person. Removing a name and making something genuinely safe to share are not automatically the same action.
Why this is a vocabulary problem as well as a digital problem
A student sees an Upload button and thinks, “I can upload this, so I may upload it.” Yet can describes capability, while may asks about permission. A single modal verb changes the meaning of the decision. A publicly accessible document might still contain private information or material that the student has no authority to redistribute.
In Singapore, MOE’s 2026 education announcements explain that AI literacy and checking generative-AI information form part of digital learning. That broad educational direction does not mean every consumer AI app or every student-data upload has school approval. The responsible question is what a specific school and tool actually permit.
Six important terms before any upload
- Public: available to a wide audience. Public visibility does not automatically erase copyright or other responsibilities.
- Private: limited to an intended audience. A private account, private message and private file may have different rules.
- Personal data: information concerning an identifiable person, sometimes through several clues taken together.
- Confidential: entrusted information whose disclosure is restricted, such as a private student report or unpublished assessment material.
- Permission: authority for a particular action from the relevant person or organisation; classroom access is not automatically permission for third-party uploading.
- Consent: agreement connected to a defined use; legal requirements and exceptions depend on context.
These are practical explanations rather than legal definitions covering every circumstance. Organisations should consult Singapore’s Personal Data Protection Commission guidance and their own policies. Students should follow their school’s authorised tools and use rules.
A science group-project decision
Imagine three fictional Secondary 2 students investigating how shade affects soil temperature. Their study goal is to make the axes of a graph clearer. Four possible inputs are available.
- A fabricated table of example temperatures, labelled “Trial 1”, “Trial 2” and “Trial 3”.
- The original class spreadsheet containing student names and mobile numbers.
- A screenshot of private feedback sent by a teacher to another student.
- A teacher-approved extract of the science question with an invented table.
The first or fourth option may meet the learning need without sharing unnecessary personal information. The second and third introduce other people’s information or entrusted material. These are teaching examples, not reports of real student behaviour.
Even changing the names in the original class spreadsheet to “Student A” and “Student B” might be insufficient if location, dates, unique observations or comments still identify a person. The safer question is whether any real personal details are needed to explain graph axes. Usually, they are not.
Preserve the learning goal while changing the prompt
Suppose a student wants writing feedback. An unsuitable prompt might be: “Here is my classmate’s marked essay and the teacher’s private comments. Tell me why she lost marks.” The difficulty is the unapproved sharing, not the subject being English.
A better prompt is: “Explain the difference between a vague thesis statement and a precise thesis statement. Invent a short example about public transport. Ask me to rewrite my own sentence.” This keeps the educational goal, but requires no identifiable classmate material.
For Mathematics, replace a genuine class contact list with an invented numerical scenario: 12 out of 30 fictional pupils choose option A. Ask how to express that as a percentage, then independently verify that 12 ÷ 30 × 100 = 40%.
For Science, a student could ask for a newly generated example of a hypothesis, controlled variable and measured variable. A teacher-approved apparatus diagram can help when necessary. Nobody needs to upload classmates’ faces to practise explaining a fair test.
Words on the AI settings screen are not interchangeable
Some systems label controls private chat, temporary conversation, history off, workspace or not used for training. These terms may concern different matters: visibility, retention, account access, use in model improvement or organisational controls. No single label answers every question.
Two extreme claims are unreliable: “Every AI tool trains on everything you enter,” and “Private mode makes any document safe to paste.” Actual settings, contracts and policies vary by provider, account and organisation. A student’s judgement should be based on the authorised service and stated conditions, not hearsay.
Learn the collocations, not just the isolated words: personal information, confidential attachment, explicit permission, authorised tool, privacy setting and external sharing. The phrase “authorised for classroom reading” is narrower than “authorised for external AI upload.”
Three checks that take less than a minute
- Whose material is it? Distinguish my own example from a classmate’s work, a teacher’s restricted paper and school records.
- Does the tool need this detail? Prefer an invented example; omit irrelevant identifiers, faces, contact details and context.
- Is this exact use approved? Check school guidance and the tool’s settings, or ask a responsible teacher.
If any answer remains doubtful, reformulate the question rather than upload the file. Say what concept you want explained, supply imaginary values and ask for a new worked example.
The distinction matters beyond school. An intern can ask an AI tool to improve a fictional customer-service email without disclosing the employer’s client list. A parent can ask for an explanation of the word “prerequisite” without pasting another family’s confidential correspondence.
Ten-minute practice: turn vocabulary into judgement
First, write six fictional cards: a public science chart; a friend’s private report; an invented Maths table; a parent email; a teacher-approved public paragraph; and a school contact list. Sort them into “generally suitable as an example,” “check authorisation,” and “do not upload without explicit approved arrangements.” Discuss why context matters.
Next, give the learner a deliberately risky sentence: “Analyse this confidential school report.” Ask them to write a prompt that preserves the educational question but uses an invented situation. Encourage the words hypothetical, fictional, general explanation and without personal details.
Then ask: “The shared folder link opens on my laptop. Does that prove its contents are public?” The student should say no, and ask about intended audience and the authority to share.
Finally, remove the cards and ask for the difference between technically possible and authorised. Revisit the distinction several days later. Knowing the word on Monday is less useful than recalling the right decision on Friday.
Teacher, parent and student actions
Teachers can demonstrate one unsafe and one safer fictional prompt, say which AI services the class is permitted to use and make the limits clear. They can teach students to request explanations, not outsource every task. Parents can use family conversations about sharing photographs and school messages to reinforce the same vocabulary. Students can make a habit of testing a prompt with imaginary details before inserting any actual material.
Privacy is separate from academic integrity. A text can contain no personal details and still be plagiarised. A student’s own work can be inappropriate to submit to AI if the assessment rules require independent completion. Teaching privacy, permission, original work and attribution as separate terms avoids one-size-fits-all advice.
A separate eduKateSG guide, How Students Check AI Tutor Explanations, addresses whether a fluent answer proves learning. That is a different question from whether the information in the prompt may responsibly be shared.
Questions learners ask
Can I simply remove a classmate’s name? Sometimes removing identifying data reduces risk, but it does not establish that the material is anonymous or authorised to share. Consider the context and whether the original is needed at all.
Does a public webpage mean I can paste its contents into any service? No. Public visibility and unrestricted redistribution are different ideas. Consider the purpose, rights, policies and whether a short question or source link will suffice.
If my teacher approves an AI tool, can it receive any file? Not automatically. Approval may be limited to particular accounts, activities and data types. Use the stated scope.
What is the best proof that a student understands these words? Give a new scenario and ask them to identify whose data it is, what the tool needs and what authorisation would be required. A correct action with a sound explanation shows more than memorised definitions.
A useful habit for the 21st century
The strongest skill is not fear of technology. It is reading words precisely enough to protect trust while remaining curious. Before uploading, ask: Whose information? Why is it needed? Who has allowed this use? Those questions can become part of responsible independent learning.
Further reading: PDPC guidance by topic and MOE 2026 announcements. This is educational guidance, not legal advice or a statement of any school’s exact AI policy.